Making heads or tails of mitochondrial membranes in longevity and aging: a role for comparative studies

Over a century ago, Max Rubner observed for six animal species that larger animals had a slower metabolic rate per unit mass and a longer lifespan compared with smaller animals. Rubner [1] Later work by Kleiber [2] and others in the 1930s supported this finding for a larger range of species. This led to several hypotheses suggesting that aging and longevity are processes that are regulated by metabolic rate.

Raymond Pearl suggested that animal tissues had a finite number of chemical reactions available, which on exhaustion led to mortality [3]. Therefore, organisms with a higher metabolism per unit mass would age and die sooner. This became known as the ‘rate of living hypothesis’.

As scientists were just beginning to understand free radical biology in the 1950s [4], Denham Harman suggested a mechanism linking metabolic rate to aging and lifespan [5]. He proposed that reactive oxygen species, being the products of metabolism, would cause cumulative damage and result in aging followed by death. This ‘free radical hypothesis of aging’ actually echoed suggestions made earlier in the century by Elie Metchnikoff that ‘senility’ may be a consequence of ‘waste’ products of metabolism [6].

Studies showing that metabolic rate-matched [7] or size-matched animals had different lifespans [7, 8] undermined the rate of living hypothesis and suggested that metabolic rate is not the exclusive determinant of lifespan. However, these early observations contributed to the question of why metabolic rate varies substantially across species, especially between size-matched endotherms (higher metabolic rate) and ectotherms (lower metabolic rate) [7]. Brand and colleagues examined metabolic rate differences in hepatocytes isolated from a mammal (a rat) and a reptile (a lizard) [9], and found that the respiration rate was fivefold higher in rat hepatocytes, possibly due to an increased amount of n-3 polyunsaturated fatty acid (PUFA) in the mitochondrial membranes [9]. However, they noted no difference in the percentage of respiration rate dedicated to processes such ATP production, proton leak across the mitochondrial inner membrane and maintenance of Na/K antiporter activity at the plasma membrane [9]. The variation in amplitude but not distribution of metabolic rate across species and its correlation with mitochondrial phospholipid composition [10], led Hulbert and Else to propose that membrane composition acts as a ‘pacemaker for metabolism’ [11]. Specifically, they postulated that membrane polyunsaturation, higher in the tissues of mammals in comparison with reptiles, would increase the molecular activity of membrane proteins thereby increasing cellular metabolic activity. Although this hypothesis held true between some species, it did not when birds were introduced into the equation, as birds have an increased metabolic rate compared to mammals, but lower membrane polyunsaturation [12].

Because of the broad but not perfect correlations of membrane fatty acid levels with metabolism, and metabolism with lifespan, a natural line of investigation developed looking at membrane composition with respect to lifespan, thus developing into the ‘homeoviscous-longevity adaptation’ [13] and later, the ‘membrane pacemaker hypothesis of aging’ [14]. These hypotheses linked membrane fatty acid unsaturation to susceptibility to oxidative damage, the propagation of which is associated with aging and mortality. In light of an increasing number of studies that support and conflict with these hypotheses, our review seeks to explore the evidence for the link between mitochondrial phospholipid and fatty acid composition, metabolism and lifespan. We discuss the roles for allometric (body size) and phylogenetic (species relatedness) corrections when making comparisons between different species [15, 16].

Membrane lipids can broadly be classified as glycerophospholipids, sphingolipids or sterols. These lipid moieties may be complexed to sugars and proteins in a cell membrane. The vast majority of mitochondrial membranes are composed of glycerophospholipids [20], which contain a glycerol backbone, a hydrophilic head group and fatty acid chains (Figure 1A). Naturally occurring fatty acids typically contain 4 to 28 aliphatic carbons of variable length and saturation: saturates contain no carbon double bonds, monounsaturates contain one double bond and polyunsaturates more than one. Figure 1B illustrates fatty acid structure and nomenclature.

Extensive work by Daum and colleagues [21, 22] has shown that the mitochondrial inner membrane is composed of all the major classes of membrane phospholipids, including phosphatidylcholine, phosphatidylethanolamine, phosphatidylinositol, phosphatidylserine, phosphatidic acid, phosphatidylglycerol and cardiolipin (CL) [22] (Table 1). Mitochondria contain a few other membrane lipids such as sphingolipids and sterols [23], the notable exception being mitochondria involved in steroid synthesis [24].

The different classes of phospholipids and fatty acids confer different properties on the membrane including its ultrastructure. As shown in Figure 2, when the diameters of the hydrophilic head groups and fatty acid chains are similar, the phospholipid molecules take on a cylindrical shape that renders the molecule suitable for forming lipid bilayers. However, small hydrophilic head groups combined with large hydrophobic fatty acid chain diameters lead to a conical shape. This favours a negative curvature, which in vitro forms hexagonal phase structures, but which in vivo is likely to store curvature stress resulting in packing defects and differential lateral pressure profiles, which may affect protein function (reviewed in [25]). Curvature stress energy can affect the binding of membrane proteins within the lipid bilayer or supply energy for protein conformational changes [25]. This is particularly important for the mitochondrial phospholipid CL, whose role in metabolism and lifespan is reviewed later.

Membrane phospholipid and fatty acid compositions are influenced by diet, which can alter membrane composition by several per cent [26]. However, there are much larger species- and tissue-specific differences in fatty acid composition, suggesting an overriding and larger effect of gene expression [15]. Indeed, the fatty acid composition of mitochondrial phospholipids varies widely across species [20] and correlates with body size, basal metabolic rate and longevity [27‐29].

Figure 3 illustrates the coupling between substrate oxidation and ejection of protons by the electron transport chain from the matrix side to the intermembrane space, thus generating a protonmotive force [36]. This electrochemical gradient can then be used to drive energy (ATP) production through ATP synthase [37]. However, electrochemical transduction is not perfectly coupled [38] and protons can leak back from the intermembrane space to the matrix via various processes including passive gradient-dependent cycling carried out by membrane fatty acids or directly by activation of proteins such as the mitochondrial uncoupling proteins (UCPs) [39]. This is termed proton leak, or uncoupling.

Brookes and colleagues [40] have shown that in simplified liposome systems from the phospholipids of eight vertebrates, representing a tenfold range of mitochondrial proton leak and a threefold difference in membrane unsaturation, the mitochondrial proton leak was similar. In a subsequent study on isolated mitochondria [10], they showed that proton leak (per milligram of mitochondrial protein) correlated with increased membrane unsaturation. Conversely, a low proton leak was associated with decreased metabolism and increased monounsaturates in the membrane. Thus, Brookes et al. concluded that mitochondrial fatty acid composition might affect the behaviour of one or more mitochondrial inner membrane proteins and thereby might affect the proton leak [10]. Furthermore, the proton leak via the lipid portion of the mitochondrial inner membrane was estimated to be only 5% of the total membrane proton leak, again suggesting that fatty acid composition might influence the proton leak via proteins, but was not the primary mediator of the process [41]. There is now good evidence that both fatty acids, especially polyunsaturates [42], and lipid peroxidation products [43] activate uncoupling proteins. The activation of uncoupling proteins by products of reactive oxygen species is thought to act as a negative feedback loop to decrease production of such species [44]. By consuming and lowering the protonmotive force, uncoupling decreases the steady-state concentration of carriers that are likely to donate an electron to oxygen to generate ROS [39, 43].

An exceptional finding to the membrane pacemaker hypothesis of metabolism is that of birds, which have a higher metabolic rate and generally live longer than size-matched mammals. One might suppose that this can be explained through mild uncoupling in birds, which would increase metabolic rate, but decrease ROS production, thus potentially explaining their longevity. However, studies have shown conflicting results in proton leak rates [10] or ROS production [8, 16, 45] in birds compared with size-matched mammals. The question of whether membrane lipids are directly correlative with uncoupling in mediating lifespan extension remains unanswered. Combining studies looking at membrane composition and uncoupling [46, 47], and membrane composition and lifespan [14, 38] requires unsafe assumptions that result in conflicting outcomes. The topic of uncoupling and lifespan is extensively reviewed elsewhere [48].

Buffenstein and colleagues, approached the question as to whether damage generation underlies species longevity by comparing oxidative damage in a long-lived rodent, the naked mole rat (MLSP >28 years) with the comparably sized mouse (MLSP 3.5 years) [61‐63]. Surprisingly and contradicting the oxidative stress hypothesis of aging, concentrations of markers of DNA damage and lipid peroxidation were greater in naked mole rats than in mice, even at a young age [62]. This is in line with data that show greater hydrogen peroxide production than expected from naked mole rat mitochondria [64]. Furthermore, contrary to predictions that oxidative stress increases with aging within species, lipid damage levels did not change with age in naked mole rats [62].

Interestingly, naked mole rats do have a membrane composition that fits with the aforementioned theoretical predictions about lifespan [65]. Compared with mice, naked mole rats have one-ninth the content of highly unsaturated DHA, despite maintaining the same overall phospholipid content [65]. Mitchell and colleagues [65] postulate that this lowers their susceptibility to peroxidative damage and state that the original findings for higher levels of lipid peroxides were because the urinary isoprostanes and liver malonaldehyde (MDA) measured in the Andziak study [62] were specific products of arachidonic acid (C20:4n-6) but not of the more unsaturated DHA (C22:6n-3) [62, 65]. Furthermore, Mitchell et al. found increased plasmenyl lipid levels for the longer-lived naked mole rats compared to mice [65], and postulated, based on previous studies, that they may act as membrane antioxidants [66, 67], so explaining the longer lifespan in these species.

However, whilst we would agree that urinary isoprostanes are products of esterified arachidonic acid, MDA is a known product of both arachidonic acid and DHA [59] and reasonably reflects lipid damage in naked mole rats. Additionally, Mitchell and colleagues do not attempt to explain why, if reduced DHA and increased levels of plasmenyl lipids in mole rats provide a protective mechanism against oxidative damage, these animals have increased mitochondrial and nuclear DNA damage as well as increased MDA levels.

Interestingly, the Mitchell study used assumptions based on previous work, which showed that only four fatty acid species are de novo synthesized whilst the rest are remodelled by enzymatic deacylation-reacylation [68]. They demonstrated that for naked mole rats compared to mice, the relative balance of fatty acids is shifted away from de novo synthesis and towards remodelling [65]. Supposing the assumptions apply correctly, this may reflect a system compensation for high oxidative stress levels, just as Andziak’s work has demonstrated that peroxiredoxin (an important antioxidant) in naked mole rats may suffer high levels of damage in keeping with its specific function [61]. Similarly, increased levels of plasmenyl lipids [65] may be a compensatory mechanism for high oxidative stress rather than a causative link with longevity. Correlations between levels of plasmenyl lipids and lifespan have not been investigated elsewhere and it would be interesting to conduct this work for a broader range of species.

The extent to which dietary PUFAs influence mitochondrial membrane phospholipids was first addressed for deer mice [55], chipmunks [53] and golden-mantled ground squirrels [69]. These studies were designed to identify the role of dietary PUFAs on torpor patterns and hibernation, and revealed that dietary PUFAs (for example, supplementary C18:2n-6 or C18:3n-3) led to a 7% increase in mitochondrial PUFA content and that these changes were paralleled by a 2.5°C decrease in minimum body temperature and longer torpor bouts [70, 71]. The duration and extent of hypothermic phases were improved by PUFAs through establishing and maintaining high membrane fluidity [72] and lowering enzyme activity, for example, for cytochrome c oxidase [73]. In other words, increased levels of PUFAs allowed for slowed metabolism and reportedly, had very beneficial effects on the survival of the animals [74].

At the time, however, scientists largely overlooked the predictions from the membrane pacemaker hypothesis of metabolism and aging and left unnoticed the fact that membrane unsaturation or PUFA contents in membranes of different tissues consistently increase in all species observed when an animal becomes torpid and lowers its metabolism (cf the membrane pacemaker hypothesis of metabolism). It took two more decades before Gerson et al.[75] compared mitochondrial metabolism between torpid and euthermic 13-lined ground squirrels. They observed that during hibernation, respiration and proton leak were suppressed as expected [75]. Unexpectedly given the lower respiration, membrane unsaturation increased while the animal was torpid and lipid peroxidative damage increased twofold as assessed by MDA levels in isolated liver mitochondria [75]. Although in a subsequent study by the same group and using the same species, mitochondrial ROS production appeared to decrease during hibernation, the assay used in that study detected hydrogen peroxide in the cytosol rather than other free radicals produced intra-mitochondrially [76]. Thus, while the pattern of increasing membrane unsaturation in the course of hibernation is consistent [77], lipid peroxidation during hibernation still remains a matter of debate.

Unlike other membrane lipids, CL is a dimerically cross-linked phospholipid that, in eukaryotes, is found almost exclusively in mitochondria and almost entirely in their inner membrane [22] (Table 1). This makes it interesting to investigate in terms of the link between mitochondrial membranes and longevity.

Because of its unique dimeric structure, CL has two glycerol backbones each with a chiral centre and four fatty acid chains, making the potential for complexity rather large (Figure 2). In eukaryotic tissues ranging from fungi to mammals, CLs contain mainly monounsaturated or di-unsaturated chains with 16 or 18 carbon atoms. This restricted fatty acid chain length and saturation result in a relatively homogeneous distribution of double bonds and carbon numbers among the four acyl chains [78].

In the mitochondrial inner membrane, CL is involved in stabilising membrane proteins including respiratory complexes [79] and the adenine nucleotide transferase [80]. Furthermore, studies show that CL directly influences the function of the adenine nucleotide transferase [81], an important mitochondrial enzyme that allows the import of ADP into mitochondria for ATP synthesis, and ejects synthesized ATP for use in intracellular processes. In the mitochondrial outer membrane, CL has been suggested to be present in and to be implicated in the function of the protein import machinery of mitochondria (reviewed in [82]). It has also been shown to have a role in regulating apoptosis through several mechanisms including interaction with caspase 8 [83] and cytochrome c [84], as well as playing a vital role in mitochondrial network morphology through interaction with fission/fusion proteins in the outer membrane (reviewed in [82]).

Despite CL’s physiological importance and its partial susceptibility to oxidative damage due to the presence of four unsaturated fatty acid chains, there is weak evidence that CL itself impairs or promotes longevity.

Many studies have used methodological approaches that provide mechanistic insights and possibly allow the authors to comment on CL’s putative role in ‘aging’ but not in lifespan [85]. For example, the response of young and aged mitochondria to exogenously supplemented CL cannot address the role of CL in lifespan [86].

At best, one yeast study showed that impaired CL synthesis lead to decreased longevity, which was restored by enhancing the stress response pathways and promoting cellular integrity using an osmotic stabiliser [87]. Although certain studies showed decreased CL levels in aged worms [88], this was consistent with their finding of decreased mitochondrial numbers and hence membranes. Interestingly, for aged rats, there is some evidence that CL fatty acid chains are remodelled from linoleic acid (18:2n-6) to the more unsaturated arachidonic (20:4n-6) and docosahexaenoic (22:6n-3) acids [89]. There is evidence elsewhere that remodelling occurs in other phospholipid species. In pulse-label experiments of phosphatidylcholine and phosphatidylethanolamine, Schmid et al. showed that only four fatty acid species were de novo synthesized (6:0–18:2 (n-6), 16:0–18:1, 16:0–22:6 (n-3) and 18:1–18:2 (n-6)), whilst the remainder were remodelled via rapid deacylation-reacylation [68]. This may explain why in a recent phylogenomic study by Jobson [90] examining codon evolution across 25 mammalian species with different longevities, of genes with significantly high evolutionary selection in long-lived species there were a number of lipid membrane composition genes. These were fatty acid elongases, desaturases and fatty acid synthases including those involved in the reconstruction of membrane CLs [90]. Again, these studies may echo our previous suggestion than PUFA levels are a response to cellular stress rather than being a causative agent in aging.